This isn’t a one time process, it has to be regular and recurring as historical vulnerabilities are discovered all of the time.
1 Like
Maybe we are thinking about this the wrong way.
Perhaps in the near future, the AI development method means an app gets recreated in toto incorporating the latest improvements to the AI’s KB/training versus the current software development methodology of creating an app and then fixing/patching/upgrading the code in perpetuity?
Instead of fixing the app, we fix iterate over the prompt/context that is used to spawn the app. Perhaps getting to the same result - improved app, increased functionality, better security, etc. but achieved through a very different “software development lifecycle” than the classic methodology now accepted and not questioned?
1 Like
Yes, certainly, agreed. Sounds like maybe a subscription to the security checker is needed. Or, perhaps it will be like SpivR is speculating, a whole new approach.
I hadn’t thought of that. That would concern me though that because of the on deterministic nature of AI, consistency would not be guaranteed. the build method could create a completely different outcome (no matter what the instructions are) Your interface could significantly change regularly in a revolution v evolution way.
What everyone is discussing sounds a lot like constantly evolving software that responds quickly to problems, vulnerabilities, and changing needs. Seems like I’ve heard of something like that.
2 Likes
So a LOT of good observations here. Security and such is a big thing. I use Monday in my jobby job so no replacing that for now. I will say that I’m using Supabase and it’s built-in database security on the backend. I’ve secured it via a password. I deployed through Netlify and secured the entire site with Google login. I’m no security expert but my guess is this is a start.
The code running on your Mac or iPhone is just the tip of the iceberg. Most robust commercial software has a big server side component that really makes the software useful. Looking at my list of apps on the iPhone, most if not all of them are useless if the iPhone isn’t connected to the internet. Let’s take Apple Maps as an example. Maybe you could “vibe code” the actual app itself, but that is probably the easiest part. What about the server side software that provides the up to date map information? Taking photos of all those roads throughout the world and storing all those terabytes of data? Scaling up and down as usage patterns change throughout the day. Keeping up with all the stores and restaurants that go out of business.
I work in the financial world and collecting and verifying financial data is a complex business. Maybe you scrape a website for data, but that website changes their format without warning and you need a way to detect that. There are also tons of regulations that are constantly changing and being updated. Will your software be in compliance with those regulations? Can you prove it in an audit?
There is just so much that goes into writing enterprise software that most people have no idea what goes on behind the scenes.
3 Likes
I agree, enterprise SaaS software isn’t dead. Apps that enable large teams to collaborate, for example, require the convergence of many disciplines to build and maintain. Just because individuals can create their own apps through natural language instruction—perhaps the highest of all abstraction layers—doesn’t eliminate the need for whole categories of applications used by teams.
Jason Fried has some thoughts on this topic that I think are well reasoned.
To paraphrase Mark Twain: I believe the reports of the death of SaaS software are a gross exaggeration 
5 Likes
In a simpler analogy, macros didn’t eliminate programming, scripts didn’t eliminate apps, and I don’t think AI will eliminate software development.
Each of these advancements has changed how software development is done, and how non-programmer users are empowered, but it is evolutionary, not revolutionary when viewed in hindsight.
I do feel there are significant disruptions on the horizon due to AI, but the specific skill of programming will not be eliminated.
3 Likes
Loved the article! I would also add economics here. While license costs sometimes feel high they quickly pale compared to the costs if the man hours to maintain a custom piece of software with all the edge case handling.
My hope for AI is simply better software. Although that is a slim hope.
3 Likes
Today’s episode of the Odd Lots podcast has a very interesting interview with Goldman Sachs CIO Marco Argenti that goes into some depth on the implications of agentic AI for software development and deployment at the enterprise level—in this case, a large, heavily regulated multinational that serves a diverse client base across a broad portfolio of products. Argenti is very a clear and articulate interviewee, and it’s worth a listen to get a picture of how large organizations with their own in-house developers are grappling with AI. A transcript is available on the podcast’s webpage.
Goldman CIO Marco Argenti on the Warp-Speed Improvements in AI